Measuring and Modeling Emission Effects for Toll Facilities Margarida C. Coelho, Tiago L. Farias, and Nagui M. Rouphail pollution levels in pay tolls. ETC is an automated system that allows drivers to pay tolls without stopping or by slightly slowing down. A typical ETC system works as follows. A transponder placed inside the vehicle is activated each time the vehicle passes a roadside sensor. The tolling agency maintains an account for each vehicle, which is debited with each use of the toll facility. Another variation uses a “smart card” of a certain value placed inside the transponder. Each time the vehicle passes a roadside reader the appropriate fee is debited from the card account. Portugal was the first country in the world to implement an integrated network of ETC systems for pay tolls (2). Currently, the system includes 90 toll plazas in the country, covering pay toll freeways, highways, bridges, parking lots, and vehicles’ pricing schemes to access specific zones in cities (2). In this paper, an approach based on experimental measurements and on the modeling of traffic and emission performance of toll facilities is presented. A mesoscopic traffic model approach is proposed for this study. It uses stochastic queuing theory (3) to estimate the queue length. Pollutant emissions were estimated by using a modal emissions approach, based on onboard measurements of vehicle emissions by driving mode (4, 5). Using the speed profile of vehicles in pay tolls, onboard emission measurements were carried out to obtain relationships between vehicle dynamics and emissions for each driving mode. The combination of the traffic and emission estimation models provides an overall pollution estimate for a toll facility under any control configuration and traffic demand patterns. The region of influence of the toll plaza was defined as the sum of the deceleration distance from cruise speed, maximum queue distance, and acceleration distance to cruise speed. The experimental data for validating the numerical traffic model were gathered on pay tolls located in three main corridors that access the city of Lisbon, Portugal. The authors are not aware of any current method that can quantify the impact of toll facilities on traffic and emission performance based on observed stop-and-go behavior and measured emission rates. Thus, the objectives of this research are as follows:

At conventional pay tolls, vehicles joining a queue must come to a stop and undergo several stop-and-go cycles until payment is completed. As a result, emissions increase because of excessive delays, queuing, and speed change cycles for approaching traffic. The main objective of this research is to quantify traffic and emission impacts of toll facilities in urban corridors. As a result of experimental measurements of traffic and emissions, the impact of traffic and emission performance of conventional and electronic toll facilities is presented. The approach attempts to explain the interaction between toll system operational variables (traffic demand, service time, and service type) and system performance variables (stops, queue length, and emissions). The experimental data for validating the numerical traffic model were gathered on pay tolls located in three main corridors that access the city of Lisbon, Portugal. The emissions model is based on real-world onboard measurements of vehicle emissions. With the appropriate speed profiles of vehicles in pay tolls, onboard emission measurements were carried out to quantify the relationships between vehicle dynamics and emissions. The main conclusion of this work is that there are two different types of stop-and-go driving cycles for vehicles joining the queue at a conventional toll booth: short and long. The length of each cycle depends on the expected queue length at the toll booth and the frequency of each cycle directly affects the level of vehicle emissions. The greatest percentage of emissions for a vehicle that stops at a pay toll is due to its final acceleration back to cruise speed after leaving the pay toll.

Conventional toll booths located on a roadway require drivers to come to a complete stop to pay the toll. One concern about this type of tolls is that vehicle emissions will increase because of the occurrence, under certain operational conditions, of excessive delays, queue formation, and speed change cycles for approaching traffic. These occurrences could have a strong impact on congestion and air quality in the surrounding urban area. Therefore, a quantification of these effects would be of interest to environmental agencies and traffic managers (1), leading to adoption of strategies aimed at minimizing traffic congestion and environmental impacts of toll collection facilities. Electronic toll collection (ETC) has been touted as one of the promising technologies aimed at decreasing traffic congestion and

1. To quantify traffic and emission impacts of toll facilities in urban corridors; 2. To develop a methodology that can quantify the traffic performance at a pay toll facility (with conventional payment and ETC), especially those parameters related to stop-and-go behavior that may have an influence on the added emissions and fuel consumption— through carbon dioxide (CO2) emissions—in the system; 3. To explain the relationship between variables characterizing stop-and-go behavior (e.g., length and number of stop-and-go cycles) with environmental and traffic performance variables—in particular, carbon monoxide (CO), nitric oxide (NO), hydrocarbons (HC), and CO2 emissions and queue length; and 4. To explain the interaction between toll system operational variables (traffic demand, service time, and number of booths) and system performance variables (stops, queue length, and emissions).

M. C. Coelho and T. L. Farias, Department of Mechanical Engineering, Instituto Superior Técnico, Av. Rovisco Pais, 1, 1049-001 Lisbon, Portugal. Current affiliation for M. C. Coelho: Department of Mechanical Engineering, University of Aveiro, University Campus of Santiago, 3810-193, Aveiro, Portugal. Current affiliation for T. L. Farias: Department of Mechanical Engineering, Superior Institute Technician, Av. Rovisco Pais, 1, 1049-001 Lisbon, Portugal. N. M. Rouphail, North Carolina State University, Campus Box 8601, Raleigh, NC 27695-8601. Transportation Research Record: Journal of the Transportation Research Board, No. 1941, Transportation Research Board of the National Academies, Washington, D.C., 2005, pp. 136–144.

136

Coelho, Farias, and Rouphail

REVIEW OF TECHNICAL LITERATURE Research in this field has focused on the development of traffic modeling tools for toll facilities and to a more limited extent on their impact on emissions. Lin and Su (6) developed a detailed methodology for the level of service analysis of main-line toll plazas, in which the operating characteristics of toll plazas are examined through field observation and simulation. Burris and Hildebrand (7) developed a discrete-event, stochastic, microsimulation computer model to determine the impact of ETC on traffic operations; submodels were developed to examine all aspects of the toll collection process, including vehicle deceleration, queuing, lane selection logistics, service time, and vehicle acceleration. Polus analyzed an urban toll plaza with a simulation model and developed a methodology for the planning of toll facilities (8–11). Al-Deek et al. (12) analyzed the improvements in traffic operations at ETC plazas. In later work, Al-Deek (13) investigated the sensitivity of the peak-hour plaza delay to market penetration of the ETC system using the toll plaza simulation model TPSIM, which is a stochastic object-oriented discrete-event microscopic simulation model for toll plazas. Astarita and Musolino (14) developed a microscopic simulation model for the evaluation of traffic impacts on toll systems; queuing theory and simulation are applied to determine the performance of a toll system and its optimal configuration as a function of traffic conditions. KLD Associates, Inc. (15) developed a commercial package entitled WATSIM (Wide Area Traffic Simulation Model) to analyze traffic operations along the approach and departure roadways for toll plazas at bridges and tunnels; it provides a stochastic integrated simulation at microscopic detail. Recent research related to the analysis and modeling of emission impacts on tolls appears to be focused mainly on the impact of ETC. Lovegrove and Wolf (16 ) discussed toll plaza design concepts that have been developed to achieve lower air pollutant concentrations while minimizing required additional right-of-way acquisition. Wang (17 ) proved the reduction of vehicle emissions using an ETC system with the MOBILE5a emissions model (18). A simple toll plaza system was studied by Lennon (19), who investigated the implementation of ETC lanes in a toll plaza and the consequences on vehicles emissions. Guensler and Washington (20) examined CO emission impacts of ETC systems compared with a normal toll plaza. Robinson and Van Aerde (21) showed how ETC could provide potential benefits to the toll systems (in terms of travel time, delay, fuel consumption, and emission levels) using the INTEGRATION microscopic simulation model (22). Sisson (23) quantified the nitrous oxide, HC, and CO production from decelerating a vehicle to 0 km/h and then reaccelerating to the same speed on a toll plaza. With the assistance of video techniques, investigation of the relationship between ETC and emissions was conducted by Lampe and Scott (24). Similar research was done by Klodzinski et al. (25), who investigated the reduction of CO, HC, and nitrogen oxide vehicle emissions at toll plazas with the implementation of ETC technology. Saka et al. (26 ) conducted a study to assess the impact of the ETC system in terms of the reduction of HC, CO, and NO emissions; the analysis involved the development of simulation and deterministic models used to generate traffic flow parameters and then to quantify emissions, using the emission models Mobile5b (18) and CMEM (27 ). The state-of-the-art review of toll facilities operation and research revealed that several models have been developed to describe traffic behavior at tolls. One of the most recognized models in this field is TPSIM (13). However, TPSIM does not currently contain a vehi-

137

cle emission estimation module. Other models on toll facilities do not take into account the additional impact of each stop-and-go cycle (each cycle is composed of a stop, acceleration, and deceleration) that a vehicle performs at a conventional pay toll for emission estimation. Instead, emission rates are calculated based on standardized driving cycles as opposed to real-world measurements (26 and references therein). The use of standardized driving cycles makes macroscopic models such as COPERT III, MOBILE6, and EMFAC— from the European Environment Agency, U.S. Environmental Protection Agency, and California Air Resources Board, respectively—less appropriate per se for evaluation of the microscale impact of traffic interruptions (1) such as pay tolls. Variability in vehicle emissions as a result of variation in vehicle operation on a toll facility can be represented and analyzed more reliably with onboard emission measurements than with the other methods such as dynamometer tests and remote sensing devices (1).

METHODOLOGY The proposed methodology consists of three steps: experimental measurements of traffic parameters, emission measurements and calculations, and implementation of the method.

Traffic Measurements To assist in the development and validation of the numerical traffic model, measurements of traffic parameters were taken at three different toll facilities in the Lisbon Metropolitan Area, in Portugal: Bridge 25th of April, Bridge Vasco da Gama (both bridges cross the Tagus River), and Carcavelos (located on Freeway A5, which connects the cities of Lisbon and Cascais). There are 16, 12, and six toll booths at each of the three sites, respectively. At the same sites six, four, and two booths operate in ETC mode, respectively. The ETC has a speed limit of 60 km/h (about 38 mph). The 1998 average annual daily traffic at the three sites is about 68,000, 19,000, and 37,000 vehicles per day, respectively (28). The Carcavelos toll facility is located on Freeway A5, which is about 30 km (19 mi) long and has two approach lanes in this zone. The maximum speed (vcruise) at the study sites is 120 km/h (75 mph). Several traffic parameters are needed to calculate the performance variables, such as approach traffic flow, service time, and vehicle dynamics. Field measurements of variables were gathered with video cameras over several days at each site. The duration of videotaping was evaluated using statistical significance tests to enable the estimation of a 95% confidence interval in relation to the average and standard deviation of the traffic stream parameters. Field measurements of the dynamic behavior of vehicles on toll facilities were also performed along each route, with the purpose of characterizing the typical stop-and-go situations, using a microwave Doppler sensor for speed and distance measurement combined with data treatment software (DLS, from DATRON-MESSTECHNIK GmbH). The times and distances of the main deceleration and acceleration were obtained from experimental measurements (5). These were considered the most detailed and most suitable to describe the acceleration and deceleration modes in toll plazas for application to emission estimation. To ensure uniformity in comparing emissions under various scenarios, the region of influence of the toll plaza was defined as the sum of the deceleration distance from cruise speed, queue length, and acceleration distance to cruise speed.

138

Transportation Research Record 1941

140

Cruise

120-60-120 kph

Cruise

120

ETC Speed (kph)

100 80 60

Conventional

40 20 0 0

20

40

60

100

80

120

140

160

180

Time (s)

Main deceleration

LSG

SSG

Idle

Main acceleration

FIGURE 1 Typical speed profile of vehicle traveling through ETC system or conventional pay tolls with stop-and-go cycles.

The characterization of stop-and-go cycles was taken from videotaping and vehicle dynamics measurements. The typical stop-and-go cycle at a conventional toll facility was characterized by measuring the elapsed time and distance traveled between successive acceleration– deceleration cycles and the corresponding number of stop-and-go situations. From the experimental measurements, it was also possible to correlate the measured queue length with the number of stopand-go situations per vehicle. It was noted that there are two types of stop-and-go situations: short (SSG) and long (LSG) (29). The long cycles represent the initial process of queue move-up times (relevant only for long queues), and short cycles represent the final process of queue move-up times when a vehicle reaches the front of the queue (payment at the toll booth). The space headway (sum of vehicle length and clear spacing between vehicles) was the unit chosen to define the threshold between these two types. Based on the work of Bell (30) and considering the average vehicle length in Portugal as well as the minimum and maximum distances between vehicles (taken from the measurements of the dynamic behavior of vehicles), a queue space headway value of 6 m (about 20 ft) per passenger car unit was chosen to distinguish between SSG and LSG. Figure 1 presents two typical speed profiles for a vehicle approaching a conven-

TABLE 1

tional toll booth and an ETC system. Table 1 gives the characteristics of the components of the typical profile on conventional toll booths (namely, the characterization of a stop-and-go unit), taken from videotaping and vehicle dynamics measurements. Figure 2 presents the relationship between the observed queue length and the number of stop-and-go cycles (total and short). It is evident that, when the queue length is short, there are very few LSG situations. On the other hand, as the queue increases, the number of LSG situations becomes significant. In addition, at very long queues, the number of SSG cycles stabilizes.

Emission Measurements and Calculations The emission estimation method adopted in this work parallels the modal emissions approach presented by Frey et al. (4), in which vehicle operation conditions are categorized into modes, such as idle, acceleration, deceleration, and cruise. Each of the categories is assumed to generate a fixed emission rate for each pollutant. Typical speed profiles of vehicles at pay tolls were gathered as the first step of the methodology. Subsequently, onboard emission measurements

Characterization of Typical Profile for Conventional Toll Booths

Typical Parameter

LSG

SSG

Maximum speed—kph (mph) Acceleration distance—m (ft) Cruise distance—m (ft) Deceleration distance—m (ft) Acceleration time—s Cruise time—s Deceleration time—s

8.6 (5.3) 4.5 (14.8) 5.3 (17.4) 4.5 (14.8) 3.8 2.3 3.8

5.4 (3.3) 1.2 (3.9) 2.6 (8.5) 1.2 (3.9) 1.5 1.7 1.6

Idle time—s

9.7

6.8

(*)Gonçalves and Farias (5). NA = not applicable.

120-0 kph(*)

0-120 kph(*)

120 (75) NA NA 300 (984) NA NA 20

120 (75) 832 (2730) NA NA 36 NA NA

NA

NA

Coelho, Farias, and Rouphail

139

Number of stop and go situations

18 Total stop and go situations (experimental)

16 Short stop and go situations (experimental)

14

2

y = -0.0239x + 1.107x - 1 2

R = 0.7375

Total stop and go situations (regression) Short stop and go situations (regression)

12 10 8 6 4

y = -0.0448x + 1.1623x - 1

2

R = 0.0039

2

R = 0.4208

0 0

FIGURE 2

2

4

6

8 10 12 Queue length (vehicles)

14

16

18

20

Queue length versus number of stop-and-go cycles.

were used to obtain relationships between vehicle dynamics and emissions for each driving mode. In this work, the average emission rates are based on onboard measurements performed by Gonçalves and Farias (5) for a EURO 4, 1.2-L gasoline-powered vehicle. The pollutants considered in this study are CO2, CO, NO, and HC. The driving modes for conventional tolls were initial deceleration from cruise speed (one mode, from 120 to 0 km/h), stop and go (two modes, LSG and SSG), idle (one mode, service time), and final acceleration to cruise speed (one mode, from 0 to 120 km/h). For ETC, one mode was considered (from 120 to 60 km/h and then again to 120 km/h). Emissions under the cruise mode (120 km/h, about 75 mph) were also considered. To reiterate, all emission comparisons were made over a reference distance that is equivalent to the sum of the main deceleration distance, queue length, and acceleration distance for a conventional toll booth. Table 2 presents the experimental characterization and classification of vehicle emissions and fuel consumption by driving mode (5). All values are given in grams per event, with the exception of the values related to the service time and cruise (g/s). The number of measurements reported by the authors (5) was 100 for each one of the stop-and-go types (short and long) and 10 for each one of the remaining modes. The measurements of fuel consumption and CO2 emissions vary between 1% (for idle) and 20% (for the remaining modes)

TABLE 2

y = 0.0378x + 6

2

2

of the average values. CO varies between 40% for acceleration and cruise modes up to 140% for deceleration, ETC cycle, and SSG. Because of the very low concentrations of CO for LSG cycles, the variability reported is 700%. Finally, for HC, it varies between 85% and 150%, while for NO, variations reported fall in the range between 50% and 200% of the average values. For conventional pay tolls, the emissions per vehicle due to main deceleration, stop-and-go situations (LSG and SSG), idle (service time), and main acceleration can be estimated as follows (Econv): Econv = EFcruise  tcruise_until_deceleration + Edeceleration + NLSG  ELSG + NSSG  ESSG + EFidle  tidle + Eacceleration

(1)

where E and EF = emissions and emission factors for driving modes, in units of g and g/s, respectively (from Table 2); NLSG = number of LSG cycles (from Figure 2); NSSG = number of SSG cycles (from Figure 2); tidle = time spent in idle mode for payment (service time) [for example, average payment times at conventional booths measured were 6.5, 8.9,

Measured Vehicle Emissions by Driving Mode Pollutant

Driving Mode

CO

NO

HC

CO2

Fuel

Main deceleration 120–0 kph

1.4 × 10 −2

Main acceleration 0–120 kph LSG cycle SSG cycle Idle—service time (*) Cruise 120 kph (*) ETC cycle 120–60–120 kph

4.0 1.1 × 10 −4 1.9 × 10 −3 5.1 × 10 −4 1.8 × 10 −2 1.4

7.3 × 10 −4 2.6 × 10 −2 9.5 × 10 −4 5.8 × 10 −4 0 (+) 1.9 × 10−4 1.1 × 10 −2

1.3 × 10 −4 1.1 × 10 −2 8.5 × 10 −4 2.9 × 10 −4 0(+) 1.9 × 10 −4 8.5 × 10 −3

8.5 245.5 17.8 12.4 6.8 × 10 −1 5.4 195.9

2.3 78.7 5.0 3.4 2.1 × 10 −1 1.6 62.2

1 mph = 1.61 kph. All values are given in grams, with the exception of the values related to the service time and cruise expressed in g s−1(*). (+) Below the detection limits of the gas analyzer.

140

tcruise_until_deceleration

Transportation Research Record 1941

and 14.4 s, respectively, for the tolls of Bridge 25th of April, Bridge Vasco da Gama, and Carcavelos (29)]; and = time spent in cruise mode required to cover the difference between the total influence area of a conventional toll plaza and the distance to perform the deceleration from cruise speed, queue length, and acceleration.

To compare emissions for conventional and ETC systems, the entire influence area of the toll facility must be considered. This includes the emissions due to the speed profile 120–60–120 km/h as well as the cruise emissions over the remaining influence area corresponding to a conventional toll: EETC = E120 − 60 −120 + EFcruise  tcruise_ETC

(2)

where E120–60–120 and EFcruise are taken from Table 2, and tcruise_ETC represents the time spent in cruise mode required to cover the difference between the total influence area of a conventional toll plaza and the sum of the deceleration distance from cruise speed (ddeceleration), maximum queue length (QL, expressed in meters), and acceleration distance to cruise speed (dacceleration) − the distance to perform the 120–60–120 km/h speed profile (d120–60–120). tcruise_ETC is computed as follows: tcruise_ETC = [( ddeceleration + QL + dacceleration ) − d120 − 60 −120 ] vcruise

(3)

where ddeceleration and dacceleration are taken from Table 1, d120–60–120 = 888 m (5), and vcruise = cruise speed (120 kph). The difference in emissions experienced at conventional pay tolls and ETC systems can then be calculated. Finally, concerning the case in which no tolls are considered, the total travel time by the vehicle is assumed at cruise speed, so the emissions are computed as follows (Eno_tolls): Eno_tolls = EFcruise  tcruise

( 4)

where tcruise represents the time spent in cruise mode required to cover the total region of influence of the toll plaza.

Implementation of the Method Because queue length appears to be a good predictor of the number and type of stop-and-go cycles for the average vehicle, implementation of the approach requires estimation of the average queue length. This can be done through direct queue length measurements or it can be estimated from stochastic queuing theory models (3) or by well-established traffic models such as TPSIM (13) taking into account the approach volume and service time. In this research, a time-dependent approach of predicting queue lengths was adopted (31). This approach considers time-dependent vehicle arrivals and service times and produces time-dependent probabilities of queue lengths at the end of each time interval in which either demand or service time varies. The profile of the expected value of queue length is then determined. Videotaping allowed for the quantification of average queue length to compare with model predictions. An example is presented in Figure 3, which represents a comparison between observed queue lengths and the time-dependent results in a 1⁄2-h period. The experimental queue lengths were observed in one of the conventional toll booths of the Carcavelos pay toll plaza, between 8 and 8:30 a.m. From experimental measurements, and with an initial queue of four vehicles, the arrival rate and service time were allowed to vary on a 1-min basis in order to be introduced in the model as inputs. The model appears to be adequate for describing the evolution of the queue length over time (the statistical correlation is R2 = .54), with a slight overestimation of queue length toward the end of the 1⁄2 h. After calculation of the queue length, the number of LSG and SSG cycles can be determined (from Figure 2). Finally, emissions are calculated with Equations 1 through 4 as described previously.

RESULTS First, the increase in unit emissions [grams per vehicle (g/veh)] associated with conventional tolls compared with ETC systems and the no-toll situation for various queue lengths (i.e., traffic congestion levels) is presented. Next, a comparison of emissions

Queue length (vehicles)

25

20

15

10

5 Numerical Experimental 0 0

200

400

600

800 1000 Time (s)

1200

1400

1600

FIGURE 3 Comparison of time-dependent observed queue length and numerical model estimates.

1800

Coelho, Farias, and Rouphail

141

for manual, ETC, and the no-toll cases is presented. This comparison focuses on the contribution of each mode to the overall emissions per vehicle due to conventional tolls. Finally, a representation of the daily emissions (g/day) for the three sites analyzed (Bridge 25th of April, Bridge Vasco da Gama, and Carcavelos) comparing conventional tolls with ETC systems and the situation of no tolls is shown. Figure 4 presents the increase in emissions (g/veh) caused by conventional tolls compared with ETC systems (Figure 4a) and a notoll situation (Figure 4b) for various queue lengths (i.e., congestion levels). The queue length represented is the average queue length over a certain time period. The main conclusion of Figure 4 is that the increased emissions caused by the presence of conventional tolls, when compared with ETC systems, are 0.015, 0.001, and 28.7 g/veh (for NO, HC, and CO2, respectively) when the queue length is one vehicle but increase to 0.022, 0.007, and 178.6 g/veh (for the same pollutants, respectively) for a queue length of 20 vehicles. The increased emissions caused by the presence of conventional tolls, compared with the situation in which there are no tolls,

is 0.020, 0.005, and 81.1 g/veh (for NO, HC, and CO2, respectively) when the queue length is one vehicle but increases to 0.028, 0.010, and 231.0 g/veh (for the same pollutants, respectively) for a queue length of 20 vehicles. CO emissions are virtually independent of queue length, because the final acceleration effect dominates all other modes (as discussed later), but CO emissions are affected by changing from ETC to conventional tolls: compared with ETC systems, the increased emissions caused by the presence of conventional tolls is equivalent to 151% for CO of the total ETC emissions (while for NO, HC, and CO2 they are 179%, 62%, and 70%, respectively), when there is a queue of 20 vehicles. A comparison of emissions between the two types of collection system (conventional versus ETC) and without the presence of tolls is presented in Table 3 (all values are given in g/veh). The percentage of each mode on the overall emissions per vehicle due to conventional tolls is also presented. For conventional tolls, an average queue length of 20 vehicles and a service time of 15 s were assumed. From Table 3, it can be concluded that the greatest percentage of emissions of a vehicle that stops at a pay toll is due to the final acceleration

CO2, CO

NO, HC 0.030

Conventional - ETC (g/veh)

250

0.025

200 NO 150

0.020

CO2 0.015

100 0.010 50

HC

0.005

CO 0.000

0 1

2

3

4

5

6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Average queue length (vehicles) (a)

CO2, CO

NO, HC

Conventional - No tolls (g/veh)

250

0.030 NO 0.025

200

CO2 0.020

150 0.015 100 HC 50

0.010 0.005

CO 0.000

0 1

2

3

4

5

6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Average queue length (vehicles) (b)

FIGURE 4 Increase in emissions (g /veh) versus average queue length for (a) conventional tolls versus ETC systems and (b) conventional tolls versus no tolls.

142

Transportation Research Record 1941

TABLE 3 Comparison of Emissions Estimates for Two Types of Collection System (Conventional and ETC) and Without Tolls for a Single Vehicle CO

Conventional Toll

Mode

NO

Value (g/veh)

% Total

HC

Value (g/veh)

% Total

CO2

Value (g/veh)

% Total

Value (g/veh)

% Total

Main deceleration 120–0 kph LSG cycle SSG cycle Idle—service time Main acceleration 0–120 kph Overall conventional

1.4 × 10−2 5.4 × 10−4 1.3 × 10−2 7.7 × 10−3 4.0 4.1

0.4 ≈0 0.3 0.2 99.1 100

7.3 × 10−4 4.6 × 10−3 3.9 × 10−3 0 2.6 × 10−2 3.5 × 10−2

2.1 13.1 11.2 ≈0 73.6 100

1.3 × 10−4 4.1 × 10−3 2.0 × 10−3 0 1.1 × 10−2 1.7 × 10−2

0.7 23.8 11.4 ≈0 64.1 100

8.5 85.7 83.6 10.1 245.5 433.4

2.0 19.8 19.3 2.3 56.6 100

Overall ETC

1.6

NA

1.3 × 10−2

NA

1.1 × 10−2

NA

254.9

NA

No tolls

6.6 × 10

NA

7.0 × 10

NA

7.1 × 10

NA

202.4

NA

−1

−3

−3

1 mph = 1.61 kph. NA—not applicable.

back to cruise speed after leaving the pay toll. The stop-and-go cycles are responsible for 39% of CO2 emissions, 35% of HC emissions, 24% of NO emissions, and less than 1% of CO emissions. The point where the effect of stop-and-go cycles overlaps the effect of acceleration is 134, 60, and 47 vehicles in the queue, for NO, HC, and CO2, respectively. Also noted is that the main acceleration is responsible for more than 99% of the CO emissions at conventional toll booths. This explains the indifference of CO emissions to queue length. Figure 5 presents the estimated daily emissions (expressed in megagrams per day) of CO and CO2 for the three sites, assuming that all daily traffic is serviced in conventional tolls (with queues of one and 20 vehicles), by ETC system or always in cruise situation (no tolls). A similar trend is noticeable for NO and HC (not presented). For CO, reductions in emissions of 61% and 84% would result if all daily traffic circulates in the ETC system and if there were no tolls, respectively. As previously discussed, CO emissions are independent of queue length, so only one column is presented for conventional tolls in this case. For CO2, emissions decrease about 41% and 53% when conventional tolls with 20 vehicles in the queue are compared with ETC systems and no tolls, respectively. CONCLUSIONS The main motivation of this research was to quantify the traffic performance and local pollutant emissions for a toll facility (with conventional payment and ETC) using real-world emission measurements and a traffic model. The parameters related to stop-andgo behavior (such as length of the queue, elapsed time between two successive acceleration–deceleration cycles, and number of stops until a vehicle departs the toll booth) were characterized. The development of this methodology involved videotaping, traffic characterization, onboard emission measurements (based on speed profiles), and emission calculation. With this methodology, a comparison of vehicle emissions for each one of the systems (conventional, ETC, or no tolls) can be performed. The main conclusions of this work can be summarized as follows: 1. On conventional pay tolls, vehicles are subjected to two types of stop-and-go situations: short and long.

2. The greatest percentage of emissions for a vehicle that stops at a pay toll is due to its final acceleration back to cruise speed after leaving the pay toll. 3. CO emissions are virtually independent of queue length, because the final acceleration effect dominates all other modes (in terms of percentages of total emissions for each mode). The final acceleration is also predominant in NO and HC emissions but to a smaller extent (e.g., for a queue length of 40 vehicles, the effect of acceleration on the fraction of total emissions decreases to 68% and 56%, for NO and HC, respectively). It is important to emphasize, however, that in terms of absolute values (total emitted grams of pollutant) longer queue lengths will produce overall higher emissions. 4. Compared with ETC systems, the increase in emissions caused by the presence of conventional tolls is equivalent to 151%, 179%, 62%, and 70% (for CO, NO, HC, and CO2, respectively) of the total ETC emissions, when there is a queue of 20 vehicles. For a queue of one vehicle, the corresponding values are 154% for CO, 117% for NO, 12% for HC, and 11% for CO2. The proposed methodology can be viewed as a decision support tool for decision makers in the following areas: optimal locations of toll facilities, reducing toll values for vehicles that use automated systems where no stopping is required, and selection of the number of open conventional toll booths to achieve balance in predefined objectives concerning service time without unduly causing excessive delay for drivers, while minimizing the deleterious effect of the stops on pollutant emissions. In areas with various geometric configurations, more measurements need to be performed to address differences in speed limits and grades. ACKNOWLEDGMENTS The research work of the first author was supported by a Ph.D. scholarship of the Portuguese Science and Technology Foundation (FCT) and European Social Fund, within the Third Framework Programme. This research had the support of FCT, Luso-American Foundation, and the U.S. National Science Foundation. The authors would like to thank OPEL–General Motors and TOTAL Portugal for providing the test vehicles and the fuel, respectively.

Coelho, Farias, and Rouphail

143

0.3 Conventional tolls

CO daily emissions (Mg/day)

ETC No tolls 0.2

0.1

0.0 Bridge 25th of April

Bridge Vasco da Gama (a)

Carcavelos

30 Conventional tolls: Queue = 20 veh CO2 daily emissions (Mg/day)

Conventional tolls: Queue = 1 veh ETC No tolls 20

10

0 Bridge 25th of April

Bridge Vasco da Gama (b)

Carcavelos

FIGURE 5 Estimated daily emissions (Mg /day) for conventional tolls, ETC, and no tolls for the three sites analyzed: (a) CO and (b) CO 2 .

REFERENCES 1. Coelho, M. C., T. L. Farias, and N. M. Rouphail. A Methodology for Modelling and Measuring Traffic and Emission Performance of Speed Control Traffic Signals. Atmospheric Environment, Vol. 39, No. 13, 2005, pp. 2,367–2,376. 2. Via Verde (Green Lane) Portugal. www.viaverde.pt. Accessed July 23, 2004. 3. May, A. Traffic Flow Fundamentals. Prentice Hall, Englewood Cliffs, N.J., 1990. 4. Frey, H., N. Rouphail, A. Unal, and J. Colyar. Emission Reduction Through Better Traffic Management: An Empirical Evaluation Based Upon On-Road Measurements. Report No. FHWA/NC/2002-001. North Carolina Department of Transportation, Raleigh, 2001. www.ncdot.org/planning/development/research/1999-08.html. Accessed July 22, 2004. 5. Gonçalves, G., and T. Farias. On-Road Measurements of Emissions and Fuel Consumption of a Gasoline Fuelled Light Duty Vehicle. Presented at VIII Clean Air Conference—Eighth International Conference on

6. 7. 8. 9. 10. 11. 12.

Technologies and Combustion for a Clean Environment, Lisbon, Portugal, 2005. Lin, F., and W. Su. Level of Service Analysis of Toll Plazas Freeway Main Lines. ASCE Journal of Transportation Engineering, Vol. 120, 1994, pp. 246–263. Burris, M., and E. Hildebrand. Using Microsimulation to Quantify the Impact of Electronic Toll Collection. ITE Journal, Vol. 66, 1996, pp. 21–24. Polus, A. Methodology and Simulation for Toll Plaza Analysis. Road and Transport Research, Vol. 5, No. 1, 1996, pp. 44–68. Polus, A., and I. Reshetnik. A New Concept and a Manual for Toll Planning. Canadian Journal of Civil Engineering, Vol. 24, No. 4, 1997, pp. 532–538. Polus, A., and J. Craus. Congestion Threshold for Lane Additions on Tollways. Transportation Quarterly, Vol. 51, No. 3, 1997, pp. 81–92. Polus, A. Dynamic Equilibrium and Concepts of Toll Plaza Planning. Traffic Engineering and Control, Vol. 39, No. 4, 1998, pp. 230–233. Al-Deek, H., A. Mohammed, and A. Radwan. Operational Benefits of Electronic Toll Collection: Case Study. ASCE Journal of Transportation Engineering, Vol. 123, No. 6, 1997, pp. 467–477.

144

13. Al-Deek, H. Analyzing Performance of ETC Plazas Using New Computer Software. ASCE Journal of Computing in Civil Engineering, Vol. 15, No. 4, 2001, pp. 309–319. 14. Astarita, V., and G. Musolino. A Microscopic Traffic Simulation Model for Toll Booth Analysis. Report CRT-99-51. Center for Research on Transportation, Montreal University, Montreal, Quebec, Canada, 1999. 15. WATSIM© (Wide Area Traffic Simulation Model). KLD Associates, Inc., Commack, N.Y. www.kldassociates.com/watsim.htm. Accessed July 23, 2004. 16. Lovegrove, A., and S. Wolf. Toll Plaza Design to Minimize Carbon Monoxide Levels at Roadway Rights-of-Way. In Transportation Research Record 1366, TRB, National Research Council, Washington, D.C., 1992, pp. 60–67. 17. Wang, S. Effect of Toll Plaza Design and Operation on System-Wide Air Pollution Emission. Master’s thesis. Department of Civil and Environmental Engineering, University of Central Florida, Orlando, 1994. 18. User’s Guide to MOBILE5 (Mobile Source Emission Factor Model). Office of Air and Radiation and Office of Mobile Sources, U.S. Environmental Protection Agency, 1994. 19. Lennon, L. Zee Bridge E-ZPass System Traffic and Environmental Studies. ITE 1994 Compendium of Technical Papers, 1994, pp. 456–459. 20. Guensler, R., and S. Washington. Carbon Monoxide Impact of Automatic Vehicle Identification Applied to Electronic Vehicle Tolling. University of California Transportation Center, Berkeley, Calif., 1994. 21. Robinson, M., and M. Van Aerde. Examining the Delay and Environmental Impacts of Toll Plaza Characteristics. Presented at VNIS/Pacific Rim Conference, Seattle, Wash., 1995. 22. Van Aerde, M. INTEGRATION User’s Guide–Volume 1: Fundamental Model Features. Queen’s University, Kingston, Ontario, Canada, 1995. 23. Sisson, M. Air Quality Benefits of Electronic Toll Collection. Transportation Quarterly, Vol. 49, 1995, pp. 93–101.

Transportation Research Record 1941

24. Lampe, A., and J. Scott. Electronic Toll Collection and Air Quality. Presented at 5th Annual Meeting of ITS America, Washington, D.C., 1995. 25. Klodzinski, J., H. M. Al-Deek, and A. E. Radwan. Evaluation of Vehicle Emissions at Electronic Toll Collection Plaza. Presented at 77th Annual Meeting of the Transportation Research Board, Washington, D.C., 1998. 26. Saka, A., D. Agboh, S. Ndiritu, and R. Glassco. Estimation of Mobile Emissions Reduction from Using Electronic Tolls. Journal of Transportation Engineering, Vol. 127, No. 4, 2001, pp. 327–333. 27. Barth, M., F. An, T. Younglove, G. Scora, C. Levine, M. Ross, and T. Wenzel. Comprehensive Modal Emissions Model (CMEM), version 2.0— User’s Guide. Center for Environmental Research and Technology, University of California, Riverside, Calif., 2000. 28. Mobilidade e Transportes na AML (Mobility and Transportation in the Metropolitan Area of Lisbon). Portuguese General Directorate of Inland Transport (DGTT)/Lisbon Transport Delegation (DTL), Lisbon, Portugal, 2000. www.dgtt.pt. Accessed July 23, 2004. 29. Coelho, M. C., T. L. Farias, and N. M. Rouphail. Assessing Traffic and Emission Impacts of Manual and Electronic Toll Facilities. Presented at 13th International Scientific Symposium Transport and Air Pollution, Boulder, Colo., 2004. 30. Bell, M. Queues at Junctions Controlled by Traffic Signals. Research Report 26. Transport Operations Research Group, University of Newcastle upon Tyne, U.K., 1977. 31. Larson, R. C., and A. R. Odoni. Urban Operations Research. PrenticeHall, Englewood Cliffs, N.J., 1981. web.mit.edu/urban_or_book/www/ book/. Accessed July 22, 2004. The Environmental Analysis in Transportation Committee sponsored publication of this paper.